基于传感器的农业温室数据直报系统与智能调控研究

Research on Sensor-based Agricultural Greenhouse Data Direct Reporting System and Intelligent Control

农业物联网、大数据、人工智能等技术的飞速发展为温室蔬菜生产的数据采集、分析、调控提供了有力支撑。为满足温室环境数据智能直报场景需求,制定了温室环境传感器的部署规范,设计并对比分析了基于LSTM、CNN-LSTM及CNN-LSTM-Attention的3种温室调控温度预测模型。基于CNN-LSTM-Attention预测模型的性能最好,其MSE、MAE、R^(2)分别为0.4570、0.3195和0.9873。设计了基于ARIMA的温室传感器数据纠错方法,实现土壤湿度预测数据值与实际测量差异不显著。整合了常见温室果蔬种植作物对环境信息的参数阈值模型,并开发了温室数据直报与智能调控系统移动端,能够指导温室环境传感器的规范部署、温度预测与纠错及常见温室果蔬种植辅助决策。研究结果为温室数据直报场景的数据采集、业务分析、温室调控提供技术手段,有助于智慧温室蔬菜产业高质量发展。

The rapid development of agricultural internet of things,big data,artificial intelligence and other technologies have provided strong support for data collection,analysis and regulation of greenhouse vegetable production.In order to meet the requirements of intelligent direct reporting scenarios for greenhouse environmental data,this study developed the deployment specifications of greenhouse environmental sensors.3 greenhouse control temperature prediction models based on LSTM,CNN-LSTM,and CNN-LSTM-Attention were designed and compared.Among them,CNN-LSTM-Attention prediction model had the best performance,with MSE,MAE and R^(2)of 0.4570,0.3195 and 0.9873,respectively.The ARIMA-based sensor data error correction method was designed and the difference between the predicted soil moisture data and the actual measurement was not significant.The parameter threshold model of environmental information of common greenhouse fruit and vegetable crops was integrated,and the mobile end of greenhouse data direct reporting and intelligent regulation system was developed.Thus,it could guide the standardized deployment of greenhouse environmental sensors,temperature prediction and error correction,and auxiliary decision-making for common greenhouse fruit and vegetable cultivation.Above results provided technical means for data collection,business analysis,and greenhouse control in greenhouse data direct reporting scenarios,and contributed to the high-quality development of the smart greenhouse vegetable industry.